The Impact of Design Effects On Standard Errors in Roadside Traffic Surveys

For the most part, random roadside traffic surveys represent complex rather than simple random samples. It is known that the design effects due to clustering, for example, often inflate the standard errors of various point estimates calculated from such surveys. This study examines the design effects in random roadside surveys conducted in three Canadian provinces in 1981. Calculating the standard errors of ratio meansforfour BA C categories, four age categories and gender in those surveys, two major conclusions were drawn. First, the calculated design effects (DEFF) were all greater than 1.0, rangingfrom 1.90 to 32.26. This implies that the standard errors of the estimates are from 1.38 to 5.68 times larger than would be obtained from simple random samples of the same size. Clustering and unequal weighting both had a major impact on increasing the overall DEFF, while countervailing stratification procedures did not reduce the effect. Second, the design effects varied considerably across the surveys even though the same basic sampling strategy was employed in all three. Further, the design effects also varied across subcategories of the same variable.